The Bonding Machine

1970 · Semi-Automatic Wire Bonder · Work 858 of 1,871

A workroom in a converted cotton mill in Lancashire, cast-iron columns and tall windows with many small panes, beneath them a row of wire bonding machines, each a dark grey steel stand on two broad side cheeks with rounded corners, a cream-white work top with rounded edges carrying a blue-grey bonding stage with a black swivel arm, brass and steel parts, hoses and a slender lamp arm, at the nearest machine a woman in a light blue overall seen from behind, bending to a stereo microscope, one hand on a lever beneath the work top, on the stage a narrow metal strip with tiny chips, grey daylight from the left.

What it was used for
Wire bonding connected the pads of the silicon chip to the leads of the package. The ASME brochure, produced by the Texas Instruments archive, gives the reason this became a production bottleneck for integrated circuits: 8 to 16 connections per chip against only 2 for a transistor; the hand technique, it says, approached an art form, and the quality of the units was highly dependent on operator skill, which varied with fatigue and distractions. Documented semi-automatic machines are TI's ABACUS, 13 of which were installed in Sherman, Texas, in 1970, and the commercially sold Multi-Matic of Ultimation in Sunnyvale: there the operator positioned the chip and pushed a button, after which the machine bonded all pads to the lead frame - 15 seconds per 14-pin package, 1 operator to 3 machines.
What was new
The process was old by 1970. Eric von Hippel records that thermocompression bonding was developed by 3 scientists at Bell Laboratories, O. L. Anderson, H. Christensen and P. Andreatch, mostly between 1955 and 1958, published in the Journal of Applied Physics in August 1957; the first commercial machine was offered by Kulicke and Soffa in late 1959, after its engineers had seen the process at a Western Electric plant in 1958 and, per the same source, made only minor mechanical modifications. Around 1970 the storing of bond positions was added: in the Multi-Matic a semiconductor read-only memory held the pad coordinates of a 14-, 16- or 18-pin device, in ABACUS a process computer. The barb: the machine was sold as a remedy against relocation - Ultimation's marketing manager David Davis reckoned that 0.5 cent per circuit beat the cost of off-shore assembly. It went otherwise: in early 1973 Electronics reported that the IC makers were still depending on offshore plants.

What the sources leave open: Which machine by which maker is meant here cannot be determined from the sources; the often-cited K&S model numbers around 1970 are not documented. The ABACUS figures come from a brochure produced by the Texas Instruments company archive, 22 years after the event.

Further reading